What is the recommended sample size for biomarker analysis?

A minimum of 30-40 samples for balanced groups (i.e. at least 15~20 each group) are recommended in order to calculate decent ROC curves and AUC evaluations.

Note: for unbalanced data, the algorithm uses Monte Carlo random sampling to produce balanced sub-samples for training data. In this case, the smaller group should have 30~40 samples

1 Like